GeT-based Ontology Construction for Semantic Disambiguation

Abstract

Semantic Ambiguation refers to the disagreement in meaning of terms used by parties in communication due to polysemy, leading to increased complexity and lesser accuracy in information integration, migration, retrieval and related activities. Semantic Disambiguation can be performed through establishing and reconciling relative positions of parties in communication by matching their terms' meanings to a common domain-specific ontology -- a knowledge representation showing concepts, organized into domains, and relationships between them. In this paper, based on the wide existence and limitations of established hierarchical ontology in a form of Classification Schemes, we proposed a novel ontology structure combining the structured nature of hierarchy with expressive capability of graphs, called Graph-embedded Tree (GeT), and a novel approach to construct a GeT-based Ontology. Evaluation was performed on United States Patent Classification System (USPC); the results showed that the information retrieval backed by GeT-based ontology yields better disambiguation capability than other typical patent search methods.